How we fetch, and how you leave
unread points people at what other people read. To do that we sometimes fetch a page you've logged — to read its title, and nothing more. Here is exactly how we behave when we do, and exactly how you take everything with you if you go.
How we fetch
When you log a link, we make a single request to that page to read its title and the name of the site. That is the whole reason we visit.
We identify ourselves honestly on every request, as unread/1.0 (+https://unread.is/how-we-fetch) — this page. If you run a site and want to know who we are, this is the answer.
One request per action you take. We never retry in a loop, and we never crawl past the page you gave us. We give up after five seconds, read at most a megabyte, and take only the title and site name — never the full text of the article. We are a link preview, not a scraper.
We do not train anything on what we fetch. There is no model here. unread delivers you to the reading; it never stands in for it.
Today we only fetch when you act. If that ever changes — if we one day check your sources on a schedule — we will respect robots.txt, and we will update this page before we do, not after.
How you leave
Your log is yours, and leaving is a feature, not a favour.
At any moment you can download everything from your settings — as a bookmarks file that opens in any browser or reader, or as complete JSON with every field we hold. One click, no waiting, no email, no ceremony.
Every profile publishes a feed that anyone can follow from any reader, with no account and no permission from us. Your readers are yours too.
If you brought your links in from somewhere else, we kept their original dates — a 2019 link still says 2019 — and we check the file before anything is brought in.
Your data leaves with you. That is the point.